Propofol impairs retinal growth and function in zebrafish
To investigate the effect of propofol on zebrafish development, zebrafish embryos were placed in an anesthetic water bath containing 2.5, 5, or 7.5 µg/ml propofol. Treatment with propofol reduced eye size at 2 dpf (Fig. 1a and 1b) in a concentration-dependent manner (83.25% for 2.5 µg/ml propofol, 90.12% for 5 µg/ml propofol, and 92.67% for 7.5 µg/ml propofol). At 3 dpf and 6 dpf, the retinas of propofol-treated zebrafish embryos had grown substantially, although the eyes remained smaller than those of controls.
Next, to clarify the effect of propofol on retinal development, we examined a series of retinal sections from propofol-treated zebrafish embryos (Fig. 1c). Histological analysis of control retinas at 2 dpf revealed well-laminated retinas with morphologically distinct ganglion cell layer (GCL), inner nuclear layer (INL), and outer nuclear layer (ONL) structures at 3 dpf and 6 dpf. In contrast, retinas from propofol-treated zebrafish embryos were poorly laminated at 2 dpf and 3 dpf, and neural layers were significantly underdeveloped, with most cells having progenitor-like appearance; at 6 dpf, the GCL could be observed, although the INL and ONL were not discernible.
To investigate the effect of propofol on retinal function, we compared the initial response of larval fish to their prey between control and propofol-treated larvae . At 6 dpf, control larvae responded to paramecia with convergent eye movements, and a high vergence angle was maintained for the duration of the hunting routine (Additional file 1). Accordingly, the extent of eye rotation during convergence was linearly associated with the initial position of the eye in control larvae (Fig. 1d). However, propofol-treated larvae exhibited abnormal behavior and lacked convergent eye movements in response to paramecia (Additional file 2), yet their motility was unaffected (Fig. 1e).
Propofol does not affect the specification of retinal progenitor cell identity
During retinogenesis, retinal progenitor cells (retinoblasts) proliferate (1). Then, beginning with the GCL and proceeding outward to the INL and ONL, they follow paths to distinct cell fates and exit the cell cycle following a terminal mitosis (2). Next, they begin to express markers of terminal differentiation (3). Finally, they undergo neuronal or glial morphogenesis (4) [33, 34].
Retinas of propofol-treated embryos lacked morphologically identifiable neurons and glia. Moreover, these cells (with the exception of some retinal ganglion cells and Müller glia) did not express terminal differentiation markers, indicating that propofol might affect retinogenesis upstream of steps (3) and (4) described above. Therefore, the loss of neurons and glia in the retinas of propofol-treated embryos could be due to a change in retinal progenitor cell or retinal cell type specification.
To assess step (1) of retinogenesis (i.e., the specification of retinal progenitor cell identity), we analyzed the expression of retinal progenitor cell identity markers (pax6a and sox2) at the optic vesicle stage (16 hpf) and the optic cup stage (26 hpf) (Fig. 3). These markers were expressed normally, indicating that propofol did not affect the specification of retinal progenitor cell identity.
Propofol affects the specification of retinal cell types
To test the third possibility (i.e., whether propofol affected the specification of retinal cell types), we first analyzed the expression of specification markers using whole-mount in situ hybridization (Fig. 4a) and real-time qPCR (Fig. 4b), including basic helix-loop-helix (bHLH) factor neurod4 and homeobox factors isl1 and crx. We found that propofol reduced the expression of these markers, suggesting that it impaired the specification of retinal cell types at an early time point. In control retinas, neurod4 expression was confined to a ring-shaped area in the INL at 40 hpf; however, in the retinas of propofol-treated embryos, neurod4 expression was reduced and limited to a much smaller patch of cells. In control retinas, crx was expressed in the presumptive ONL at 42 hpf; however, propofol-treated embryos lacked detectable crx expression. At 48 hpf, isl1 expression was weak and confined mainly to the GCL; isl1 signals were detected in the retinas of propofol-treated embryos. Expression of markers in the brain (isl1) and pineal body (crx) appeared relatively unaffected.
bHLH and homeobox factors are known to drive proliferative progenitor cells out of cell cycle and stimulate their differentiation [34, 35]. First, we used the TUNEL assay to analyze the number of apoptotic cells in the retinas of control and propofol-treated embryos at 36 hpf, 48 hpf, and 72 hpf (Fig. 5a). The number of TUNEL-positive cells was higher in the retinas of propofol-treated embryos than in control retinas at each time point, indicating an increase in cell death.
Next, we used a series of EdU incorporation assays to examine how retinoblasts exit the cell cycle. The percentage of EdU-positive cells was significantly lower in the retinas of propofol-treated embryos than in control retinas during and after specification of retinal cell types (Fig. 5b). At 36 hpf (the initial stage of specification), the retinas of propofol-treated embryos contained 51.7% fewer EdU-positive cells than control retinas (64.3%). At 48 hpf (the intermediary stage of specification), the retinas of propofol-treated embryos contained 28.9% of EdU-positive cells, limited to the peripheral retina and the ciliary marginal zones; control retinas contained 46.2% of EdU-positive cells, spread throughout the central retina. At 72 hpf (the end stage of specification), only 8.1% of EdU-positive cells were found in the retinas of propofol-treated embryos, whereas 27.2% of control retinal cells remained in the S-phase.
Taken together, these data supported the third hypothesis that laminar organization of neuronal layer would be impaired due to alterations in the specification of retinal cell types. Indeed, in propofol-treated embryos, the specification of retinal cell types was inhibited because retinoblasts exited the cell cycle too early, i.e., cell cycle progression was compromised.
Propofol impairs retinal development via inhibition of zisp expression
DHHC family-mediated palmitoylation markedly affects neural development, and deficiency or dysfunction of PAT activity is associated with several neurological disorders. We used qRT (quantitative reverse transcription)-PCR to analyze the expression of all DHHC family members in the retinas of propofol-treated embryos (Fig. 6a). We observed a significant decrease in zisp/zdhhc8 expression levels after treatment with propofol. These results indicated that propofol impaired the specification of retinal cell types possibly via inhibition of zisp expression.
Notably, zisp transcripts were hardly detected before the bud stage (10 hpf) using whole-mount in situ hybridization (Additional file 3a and 3b). At the 14-somite stage (16 hpf), in addition to the somitic expression, zisp transcripts were found in the optic vesicle (Additional file 3c and 3d) . At 24 hpf, zisp was detected in the retina and lens (Additional file 3e). In the retina, zisp expression was restricted to the INL at 32 hpf (Additional file 3f) and 48 hpf, and thereafter decreased gradually (Additional file 3 g) until it disappeared at approximately 60 hpf. RT-PCR analysis revealed that zisp was maternally expressed in zebrafish (Additional file 3 h).
There is a diverse network of intrinsic signaling pathways  and transcription factors  determining the normal specification and differentiation of multiple cell types during retinogenesis. To investigate the possible mechanism via which Zisp might be involved in retinogenesis, we overexpressed zebrafish Zisp in wild-type embryos at the one-cell stage (Fig. 6b). All embryos injected with 80–100 pg of zebrafish zisp mRNA showed a dorsalization phenotype characterized by an up-turned tail and a lack of the ventral fin, which was reminiscent of bone morphogenetic protein (BMP) signaling mutants . Increasing the dose of zisp mRNA (120–200 pg) caused a stronger phenotype with further reduction of the tail and trunk in 68% of the injected embryos. Indeed, Zisp antagonized BMP function and inhibited ventralization phenotypes and eve1 induction (a BMP target gene) by BMP ligands (bmp2b and bmp7) but not by BMP receptors (bmpr1a and bmpr1b) or Smad1 (a BMP downstream component) (Fig. 6c and 6d). Thus, Zisp likely acted upstream of BMP receptors or BMP signaling.
Zisp promotes Noggin-1 secretion and stability
Early embryos were further analyzed using in situ hybridization and RT-PCR. Following Zisp overexpression, eve1 and gata2 (another BMP target gene) expressions were strongly reduced or even abolished (Fig. 6e and 6f), whereas shh and gsc (a homeobox gene) extends all around the margin (Fig. 6g and 6 h). Notably, although bmp2b expression was unaffected at early blastula stages, it was lost in the ventral blastoderm at later mid-gastrula stages in Zisp-overexpressing embryos (Fig. 6i and 6j). This pattern is similar to the biological activity of Noggin , indicating that Zisp might regulate Noggin.
Zebrafish possess three nog paralogs: nog1, nog2, and nog3 . To investigate whether Zisp regulates Noggin and, if so, to determine which Noggin protein is regulated, we co-transfected GFP-tagged Zisp and HA-tagged Noggin in RPE1 cells. We found that Zisp significantly enhanced Noggin-1 expression, while Noggin-2 and Noggin-3 were not detected (Fig. 7a).
Structurally, Zisp is predicted to have four hydrophobic segments (probably transmembrane domains and a DHHC zinc finger motif), which are evolutionally conserved from yeasts to the human species (Additional file 4). This raises two questions: (1) is zebrafish Zisp a PAT and, if so, is Noggin-1 a substrate for palmitoylation by Zisp; and (2) what is the biological relevance of Noggin-1 palmitoylation? To answer the first question, RPE1 cells were co-transfected with HA-tagged Noggin-1 and GFP-tagged Zisp. We found a high degree of co-localization between these two proteins (Fig. 7c). Next, we examined the level of palmitoylation using the ABE assay. We found that Noggin-1 was palmitoylated by wild-type Zisp (Fig. 7c, lanes 1 and 2). However, the ability of a mutant form of Zisp (lacking the DHHC domain) to palmitoylate Noggin-1 was greatly reduced (Fig. 7c, lane 6).
In zebrafish, Noggin-1 has two potential palmitoylation sites near the C-terminus (Fig. 7b), which are conserved across species. To investigate which of the two Noggin-1 cytoplasmic cysteine residues can be palmitoylated, we mutated them to alanine residues. Mutation of the cysteine residues in various combinations revealed that not only did Zisp mediated palmitoylation most prominently at Cys-212 (p < 0.001) but also at Cys-214 (p < 0.0037). Mutation of both Cys-212 and Cys-214 led to complete inhibition of palmitoylation (p < 0.001) (Fig. 7c, lanes 3–5 and 7D). Therefore, we concluded that zebrafish Noggin-1 is palmitoylated on Cys-212 and/or Cys-214.
Notably, we found that Noggin-1 was mainly distributed in cell–cell junctions in control retinas (Fig. 7e) and within the cytoplasm in the retinas of propofol-treated embryos. Thus, propofol might impair the development of retinogenesis via direct regulation of Noggin-1 secretion and trafficking.
To investigate the biological effect of propofol on Noggin-1, we treated HA-tagged Noggin-1–transfected RPE1 cells with propofol and investigated Noggin-1 secretion and subcellular distribution (Fig. 7f). Propofol reduced the levels of secreted Noggin-1 in a time-dependent fashion (Fig. 7f, lanes 1–5). Soluble intracellular Noggin-1 was increased at low propofol doses (Fig. 7f, lanes 1 and 2) but reduced at high propofol doses (Fig. 7f, lanes 3–5). Surprisingly, when we added the proteasome inhibitor MG132, the levels of both total and soluble intracellular Noggin-1 did not change in response to propofol (Fig. 7g), indicating that upon treatment with propofol, the intracellular pool of Noggin-1 moves into the Triton-soluble fraction, which is readily degraded in a proteasome inhibitor-dependent fashion. Taken together, these results led us to raise the hypothesis that Noggin-1 palmitoylation might be required to stabilize a membrane-associated form of Noggin-1 (i.e., the Triton-insoluble fraction) that may serve as a precursor for secreted Noggin-1, and that propofol might impair this stabilization.
To test this hypothesis, we first investigated the effect of reducing or increasing Zisp levels on Noggin-1 secretion into the media. In RPE1 cells transfected with nog1 alone, Noggin-1 protein was efficiently secreted, and Noggin-1 could be detected in both Triton-soluble and -insoluble fractions (Fig. 7g, lane 2). When zisp was co-transfected with nog1, the levels of secreted Noggin-1 increased (Fig. 7g, lane 3). However, co-treatment with propofol reduced Noggin-1 secretion and the amount of Triton-insoluble Noggin-1, while not clearly affecting the levels of soluble intracellular Noggin-1 (Fig. 7g, lane 4).
Next, we co-expressed mutant forms of nog1 with zisp in COS cells, and determined the amount of secreted Noggin-1 (Fig. 7h). Although C212A, C214A, and C212A/C214A mutations resulted in only a slight change in baseline Noggin-1 secretion (Fig. 7h, lanes 1, 3, 5, and 7), they reversed the increase in Noggin-1 secretion induced by co-transfection with Zisp (Fig. 7h, lanes 2, 4, 6, and 8). In addition, we found that the level of soluble intracellular Noggin-1 was higher in cells expressing the mutant forms of nog1 (Fig. 7h, lanes 1, 3, 5, and 7). Meanwhile, in cells expressing the C212A/C214A mutant, the amount of Triton-insoluble Noggin-1 was significantly reduced (Fig. 7h, lanes 7 and 8).
Taken together, these results suggested the dual effect of Zisp: it promoted wild-type Noggin-1 secretion and stabilized membrane-associated Noggin-1. Propofol inhibited Zisp expression and Noggin-1 trafficking.